Articles: CPU
 

Bookmark and Share

(0) 

Table of Contents

Pages: [ 1 ]

When in 1997 an absolutely unknown company called IDT announced their desire to produce x86 processor it soundedabsolutely unbelievable. Everybody didn't have even the slightest doubt that four do-it-yourselfers, who gathered togetherfor a cup of coffee and suddenly decided to make such a complicated microchip without any preliminary research, will fail.And the skeptical attitude towards this bold attempt increased even more after they had revealed the first details concerningthe new product - C6. Those 4 brave guys made up their mind to assemble a Pentium-like CPU based on 486 architecture.However, they have maximally simplified it: no superscalar pipelining, no out-of-order execution, no speculative execution,no rename registers, no reordering buffers. Even branch prediction was absent! But the processor turned out rather cheap,with low power dissipation and a small core size.

Well, all this looks terrific, but as for users they usually pay attention not to the core size but to the CPUperformance, and here IDT C6 couldn't boast much. In other words it could stand the competition only in integer arithmetic,but as soon as it came to anything else arithmetical FPU as well as MMX performed twice as badly as their Intel competitors.All these drawbacks together with a completely wrong IDT marketing policy prevented C6 from being the top sales item.

However, there is a couple of cool features, which helped C6 leave everyone else far behind. The first thing is backwardcompatibility. Starting from Intel Pentium MMX and AMD K6 all Socket-7 processors began using a lower core voltage -2.8-2.9V, while a common Pentium required 3.3-3.5V. So, if you decided to upgrade your PC and shift from the old Pentiumto some new Socket-7 CPU, you would have to buy a new mainboard. Then the 100MHz FSB of new AMD K6 and K6-2 models wouldagain require an additional motherboard upgrade. And as a result - no compatibility could be achieved. This situation suitedperfectly for IDT C6, which was later called Winchip. It was a CPU that accepted the old core voltage - 3.3-3.5V - and 66MHzFSB frequency, and it allowed the users to upgrade their computers by simply installing an IDT Winchip on the oldPentium-class mainboard. The second feature of Winchip that is worth mentioning is its low power as well as a relatively lowheating compared to its competitors, which makes this processor suitable even for cheap notebooks. However, the reasonsdescribed above prevent it from being widely used here.

Having taken into account the experience IDT acquired in the CPU market, they decided to improve their Winchip and notso long ago launched a new model - Winchip 2. You are probably wondering what it looks like. Here you are. We'd like tooffer you the detailed specifications list of this processor intended for low-cost PCs:

  • Installs into Socket-7, which is electrically and physically compatible with it;
  • Software compatible with Intel Pentium family and all x86 applications;
  • MMX-compatible;
  • 64KB L1 cache - 32KB for instructions and 32KB for data;
  • 2 superscalar MMX blocks;
  • Works at 66MHz and 100MHz FSB frequency;
  • A relatively small core size - 58 mm2 with 0.25 micron technology. Besides, chip versions with 0.35 micron technologyare also available;
  • Supports 3DNow! instructions. There are 2 pipelines operating with two pairs of single precision real numbers;
  • 3DNow! is supported by DirectX 6.0 and up;
  • The following models are available: 225MHz (75MHz FSB), 233MHz (100MHz FSB), 240MHz (60MHz FSB), 250MHz (83MHzand 100MHz FSB), 266MHz (66MHz and 100MHz FSB), 300MHz (75MHz and 100MHz FSB);
  • Core voltage 3.3 or 3.5V depending on the supply.

Well, it is high time we dwelled on the differences between Winchip and its elder brother Winchip 2. Firstly,Winchip 2 got a pipelined FPU, which is very effective for processor performance improvement in the games using coprocessortogether with integer instructions, as for instance, in Quake2. Here IDT has left far behind such authority as AMD, whichfailed to provide a pipelined FPU even for its K6-3. Secondly, Winchip 2 has an additional MMX pipeline (a second one) andhence the performance of this block has doubled. In fact, this innovation seems not very justified and logical since verylittle software applies MMX-instructions. And thirdly, IDT has licensed AMD 3DNow! technology, which provides a significantperformance increase when calculating 3D-scenes parameters in case the working applications support it. And it was notwithout purpose that IDT preferred 3DNow! to Intel SSE for its CPU: the number of existing applications supporting 3DNow!is much bigger because this technology appeared in the market considerably earlier than SSE. Probably, IDT hopes to retainits superiority with 3DNow! over SSE for a while. And finally, the set of acceptable processor frequencies has been enlarged,and some models got 100MHz FSB frequency support.

In other words, the new IDT Winchip 2 still remains the best way for upgrading without changing the mainboard, no matterif there is a 100MHz FSB or not. Core voltage was left unchanged that is why BIOS update is the only necessary requirementfor booting any Socket-7 mainboard with IDT Winchip 2 on it. In fact, it is a wonderful thing for a low-cost upgrade!

Currently, there are two Winchip 2 revisions. The newer revision (A) additionally supports fractional frequencymultipliers. The previous revision of this CPU as well as the original Winchcip can multiply FSB frequency only byintegers. Now it became possible to apply not only all usual multipliers but also to multiply FSB frequency by two veryinteresting numbers: 2.33x and 2.66x, which are generated by the processor and are set on the mainboard equal to 5x and 5.5xcorrespondingly. They may be used together with a 100MHz FSB frequency.

Like Cyrix, IDT turns to PR-rating to mark its 100MHz processors. In particular, we've got a 266MHz processor fortesting, which worked at a real frequency equal to 233MHz (2.33x100MHz). To evaluate this rating IDT compares itsperformance to that of AMD K6-2 in ZD Winstone 99 and ZD 3DWinbench 99.

That's about all concerning IDT Winchip 2. And now let's pass over to the tests. In our benchmarks we will use100MHz FSB IDT Winchip 2 266 and AMD K6-2 266. Besides, we've also tested Intel Celeron 266MHz. The testing system wasconfigured as follows:

  • Chaintech 6BTM and Chaintech 5AGM2 mainboards;
  • ASUS V3400TNT graphics card (on Nvidia Riva TNT chipset) with Detonator drivers;
  • 3D-accelerator Creative 3D Blaster Voodoo2;
  • Sound card on Ensoniq ES1370;
  • IBM Titan DTTA 371010 harddisk;
  • 128MB SEC PC-100 SDRAM;
  • Windows98;
  • All 3D-tests were carried out with the resolution 800x600x16.

First of all we checked IDT Winchip 2 performance in Winstone 99, which is intended for performance evaluation inoffice applications.

Actually, it was quite predictable that AMD K6-2 266 and IDT Winchip 2 266 would show practically the same results,and our expectations came true. No wonder because PR equal to 266 was assigned to IDT Winchip 2 as a result of thisparticular test. And even earlier Winchip had also performed not bad in terms of integers. But now Winchip 2 has improvedthe performance by 10% compared to Winchip.

Now let's have a look at CPUMark99:

This test shows the performance of integer arithmetic and memory subsystem bandwidth rate. As we see, the result shownby Winchip 2 is lower than that by AMD K6-2, which is more likely to mean that Winchip 2 still has room for improvement asfar as working with memory is concerned.

Then we've got an FPU performance chart:

In comparison with IDT Winchip we hardly see any changes here: the FPU was and still is its weak point. However,the obtained result doesn't necessarily mean that the situation won't change in real applications. The thing is thatthough Winchip 2 coprocessor is not so fast as that of AMD K6-2, it operates simultaneously with the integers, which isimpossible with the non-pipelined FPU of its AMD opponent. This feature influences processor performance in realapplications, which is displayed in the chart below. By the way, the result achieved by Intel Celeron 266 given for abetter comparative analysis, clearly shows the great improvement of Intel's arithmetical FPU. To say the least of it,Intel's competitors could hardly dream of it! But let us look at the CPUs performance in a gaming application requiringa pipelined FPU. Of course, everybody recognizes Quake2! So, here you are:

Of course, 3DNow! support was disabled here so that the test could concentrate only on FPU performance. Comparing theperformance of the arithmetical FPUs only we'll see that IDT Winchip 2 is significantly lagging behind, but due to thepipelining technology applied in IDT Winchip 2 coprocessor it managed to catch up with AMD K6-2 and to show very similarresults. So, once again pipelining technology proved to be extremely cool!

Passing over to MMX block we'll see that its performance in IDT Winchip 2 has doubled compared to its predecessor -IDT Winchip. We utilized Intel Media Benchmark to test the performance of this block. This benchmark carries out MPEG-1decompressing, image processing with various digital filters and 3D-rendering based on MMX instructions. Here are theresults obtained:

Frankly speaking, the results can hardly be impressive. IDT Winchip 2 MMX is really poor. The only consolation forIDT is a relatively small number of applications utilizing MMX-instructions, which is very unlikely to strongly increasein the near future. MMX is a set of integer SIMD instructions, which are not so popular among software developers, whoprefer floating point arithmetic provided by 3DNow!. That is why the quality of 3DNow! technology in Winchip 2 turns outmuch more interesting and important. In this respect let's find out CPU 3DMark results:

Well, the situation with IDT Winchip 2 leaves much to be desired. With the enabled 3DNow! AMD K6-2 shows over 1.5 timesbetter performance than IDT Winchip 2, which actually means something. The real 3DNow! applications such as Quake2 workingthrough 3dfx OpenGL miniport demonstrate the following:

As we have expected, IDT Winchip 2 greatly falls behind. Nevertheless, it performed not that badly compared to IntelCeleron, which has hardly heard about any 3DNow!.

In conclusion, let's devote some time to processor performance in a popular game occupying the whole CPU - to Unreal.

No brilliant achievements but on the whole the situation is quite satisfactory. However, IDT Winchip 2 costs almost asmuch as AMD K6-2 of the same frequency that's is why taking into account their performance the latter will undoubtedly be thebest choice for any of you.

And now a few words about overclocking. When we found out that IDT Winchip 2 wasn't getting any warmer while working, ahope crept in that it should perform beautifully at higher frequencies. Ha! Tell us more! We suffered a great fiasco tryingto apply the widely spread overclocking method to this stubborn CPU. Even the slightest frequency increase over the nominalone led to unstable performance. So, unfortunately, we have to grieve the overclockers: there is no way to make their hopescome true.

Conclusions

Summing up, we'll get the following picture. To be objective, all tests showed that IDT Winchip 2 failed to surpassAMD K6-2, which is not much more expensive. In other words, if you aim at finding the best Socket-7 processor it isdefinitely AMD K6-2 or K6-3. However, you shouldn't forget Winchip's two main advantages. The first one is availabilityof really cheap younger models, and hence IDT retains its dominating position in the low-cost market since AMD is nolonger presented there. And the second thing is an absolutely unimpeded upgrade when you can avoid changing your mainboardfor a newer revision, which supports dual CPU voltage and 100MHz FSB frequency. So, IDT Winchip 2 proves to cope with itsmain tasks. Later on IDT is going to improve its CPUs performance at the same frequencies but with a doubled L1 cachesize.

But the main conclusion is a bit different. IDT made the world admit that it shouldn't be necessarily a giant corporation,which is capable of assembling such a complicated device as a microprocessor. A small group of skillful specialists may alsobe enough for that. And pretty soon Rise seems likely to prove this statement once again. Well, we'll see! 

Pages: [ 1 ]

Discussion

Comments currently: 0

Add your Comment




Latest materials in CPU section

Article Rating

Article Rating: 10.0000 out of 10
 
Rate this article:
Excellent
Average
Poor